HAL Id: hal-01372848
https://hal-insu.archives-ouvertes.fr/hal-01372848v2
Submitted on 26 Jun 2017
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
The influence of synoptic circulations and local processes on temperature anomalies at three French observatories.
Cheikh Dione, Fabienne Lohou, Marjolaine Chiriaco, Marie Lothon, Sophie Bastin, Jean-Luc Baray, Pascal Yiou, Aurélie Colomb
To cite this version:
Cheikh Dione, Fabienne Lohou, Marjolaine Chiriaco, Marie Lothon, Sophie Bastin, et al.. The influ- ence of synoptic circulations and local processes on temperature anomalies at three French observa- tories.. Journal of Applied Meteorology and Climatology, American Meteorological Society, 2017, 58 (1), pp.141-158. �10.1175/JAMC-D-16-0113.1�. �hal-01372848v2�
The influence of synoptic circulations and local processes on temperature
1
anomalies at three French observatories
2
Cheikh DIONE1∗†, Fabienne LOHOU1, Marjolaine CHIRIACO2, Marie LOTHON1, Sophie
3
BASTIN2, Jean-Luc BARAY3, Pascal YIOU4and Aur´elie COLOMB3
4
(1) Laboratoire d’A´erologie, Universit´e de Toulouse, CNRS, UPS, France.
5
(2) LATMOS/IPSL, UVSQ Universit´e Paris-Saclay, UPMC Univ. Paris 06, CNRS, Guyancourt, France.
6
7
(3) Laboratoire de M´et´eorologie Physique, UMR 6016 Universit´e Blaise Pascal/CNRS, Clermont Ferrand, France.
8
9
(4) Laboratoire des Sciences du Climat et de l’Environnement, UMR8212 CEA-CNRS-UVSQ, Universit´e Paris - Saclay & IPSL, Gif-Sur-Yvette, France.
10
11
∗Corresponding author address: Dr. Cheikh DIONE, Centre de Recherches Atmosph´eriques, 8 route de Lannemezan, 65300 Campistrous, France.
12
13
E-mail: Cheikh.dione@aero.obs-mip.fr
14
†Current affiliation: Laboratoire d’A´erologie, Universit´e de Toulouse, CNRS, UPS, France.
15
LaTeX File (.tex, .sty, .cls, .bst, .bib) Click here to download LaTeX File (.tex, .sty, .cls, .bst, .bib) dione_et_al_2016.tex
ABSTRACT
The relative contribution of the synoptic-scale circulations to local and mesoscale processes was quantified in terms of the variability of middle lat- itude temperature anomalies from 2003 to 2013 using meteorological vari- ables collected from three French observatories and reanalyses. Four weather regimes were defined from sea level pressure anomalies using National Center for Environmental Prediction (NCEP) reanalyses with a K-means algorithm.
No correlation was found between daily temperature anomalies and weather regimes, and the variability of temperature anomalies within each regime was large. It was therefore not possible to evaluate the effect of large scales on temperature anomalies by this method. An alternative approach was found with the use of the analogues method: the principle being that for each day of the considered time series, a set of days which had a similar large-scale 500 hPa geopotential height field within a fixed domain were considered. The ob- served temperature anomalies were then compared to those observed during the analogue days: the closer the two types of series, the greater the mark of the large scale. This method highlights a widely predominant influence of the large-scale atmospheric circulation on the temperature anomalies. It showed a potentially larger influence of the Mediterranean Sea and orographic flow on the two southern observatories. Low-level cloud radiative effects substantially modulated the variability of the daily temperature anomalies.
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
1. Introduction
36
Temperature fluctuations in France and, more generally, Western European are largely connected
37
to large-scale weather regimes. However, the processes linking atmospheric variability to surface
38
temperature may vary with the season. Cattiaux et al. (2010) used 500 hPa geopotential height to
39
define weather regimes influencing Europe and found that the cold winter of 2010 was associated
40
with a large occurrence of the negative phase of the North Atlantic Oscillation (NAO) weather
41
regime. In summer, major heat waves over France and the UK are generally linked to persistent
42
anticyclonic conditions (such as those in 2003) (Cassou et al. 2005; Yiou et al. 2008). They
43
may also be linked to Atlantic low pressure, which leads to southerly flows (such as those seen
44
in 2015), with amplification by soil moisture-temperature and boundary-layer feedbacks (Sch¨ar
45
et al. 2004; Seneviratne et al. 2006; Fischer et al. 2007b; Vautard et al. 2007; Quesada et al. 2012;
46
Miralles et al. 2014). Warm winters are linked to a zonal westerly flow (such as in 2007 and 2014)
47
(Luterbacher et al. 2007), which can be amplified by land albedo and cloud radiative effects. The
48
role of such amplifying factors was investigated mainly with regional model simulations (Zampieri
49
et al. 2009; Stegehuis et al. 2013; Seneviratne et al. 2004; Stefanon et al. 2014), but it was proven
50
necessary to use high-resolution observations to validate such an approach, since models seemed
51
to exacerbate the role of these factors over Europe (Cheruy et al. 2014; Bastin et al. 2016). For
52
example, Chiriaco et al. (2014), using a combination of space and ground-based observations and
53
twin simulations, showed that the heat wave that occurred over Northern Europe in July 2006
54
was linked to specific large-scale conditions favoring a low cloud deficit over this area and was
55
amplified by dry soil, which contributed to about 40% of the anomaly.
56
As for the weather regimes, a flow analogue method was also used to study the seasonal vari-
57
ability of surface temperature anomalies over Europe (Cattiaux et al. 2010; Chiriaco et al. 2014;
58
Vautard and Yiou 2009; Yiou et al. 2007). Cattiaux et al. (2010) found a larger positive depar-
59
ture of observed temperatures from flow analogues for minimum than for maximum temperatures.
60
They observed a maximum departure over the Alps region. Spatial variability and underestimation
61
of observed temperature anomalies by reconstructed temperature anomalies suggest an important
62
role of the smaller scale processes concerning temperature anomalies. France is located in a tran-
63
sitional region between subtropical influences and Atlantic perturbations. It covers an area where
64
climatic model predictions have suggest significant uncertainty, with large scatter in temperature
65
and precipitation due to different sensitivities to local processes (Bo´e and Terray 2014). For these
66
reasons, it is useful to employ observational data to quantify the influence of large-scale atmo-
67
spheric circulations relative to those of local processes to help explain the variability of daily
68
temperature anomalies across France.
69
Our study aims to quantify the relative contributions of large-scale atmospheric circulations and
70
of local processes on the variability of temperature anomalies at three observatories located in
71
France. For this, we shall evaluate specific issues : (i) the effect of weather regimes on daily tem-
72
perature anomalies by use of the classification of weather regimes defined from sea level pressure
73
(Yiou and Nogaj 2004), and (ii) the capability of local processes to amplify or reduce temperature
74
anomalies by use of flow analogue atmospheric circulations based on geopotential height at 500
75
hPa (Yiou et al. 2007). Our analysis is based on a series of meteorological variables (temperature,
76
wind and radiation) observed at three observatories from ROSEA (R´eseau d’Observatoires pour
77
la Surveillance de l’Eau Atmosph´erique) national network. It is also based on reanalyses (NCEP
78
and ECMWF).
79
The manuscript is organized as follows. In section 2, the three ROSEA observatories, the cor-
80
responding datasets, the large-scale diagnostic and the methodological approach are presented.
81
Section 3 presents the analysis of large-scale conditions versus local processes using flow ana-
82
logues. Conclusions appear in section 4.
83
2. Data and methodology
84
a. Observatories
85
In this study, we use surface observations from three observatories (SIRTA (Site Instrumen-
86
tal de Recherche en T´el´edetection Active), COPDD (C´ezeaux-Opme-Puy De Dˆome) and P2OA
87
(Plateforme Pyr´en´eenne de l’Observation de l’Atmosph`ere)) from the five ROSEA network ob-
88
servatories, located across varied landscapes along a North-South transect across France (Figs. 1a
89
and 1b).
90
1) SIRTA
91
The northern observatory of ROSEA is known as SIRTA (48.7◦N-2.2◦E and 160 m elevation)
92
(Haeffelin 2005). SIRTA is located on a plateau in a suburban area in Palaiseau, 20 km south-
93
west of Paris (Fig. 1a). It is dedicated to the research of physical and chemical processes in the
94
atmosphere, mainly using remote sensing. Since 2002, observations of precipitation, water va-
95
por, clouds, meteorological variables, atmospheric gases, solar radiation, and wind power have
96
been collected. More details concerning the SIRTA observatory can be found in Haeffelin (2005)
97
or on the following website: http://sirta.ipsl.polytechnique.fr/sirta.old/. Quality control and ho-
98
mogenization of the data yielding a uniform hourly time-resolution was undertaken at SIRTA for
99
the entire observation period (Chiriaco et al. 2014; Cheruy et al. 2013). This project, named
100
SIRTA-ReOBS, provides a single netCDF file with more than 40 variables from 2003 to 2013
101
(http://sirta.ipsl.polytechnique.fr/sirta.old/reobs.html).
102
2) C ´EZEAUX-COPDD
103
The COPDD observatory is located in the Auvergne region, in the center of France (Fig. 1a)
104
where various in situ and remote sensing instruments continuously measure the atmospheric dy-
105
namics, radiation, atmospheric gases, cloud microphysical variables and aerosols. This observa-
106
tory is composed of three instrumented sites: C´ezeaux (at an altitude of 394 m, an urban site,
107
Opme (at an altitude of 680 m), and Puy-De-Dˆome (at an altitude of 1465 m). In this study, we
108
use the meteorological variables collected at the C´ezeaux site in order to obtain relatively similar
109
terrain across the three sites. The C´ezeaux site (45.47◦N-3.05◦E) is located on a plain on the cam-
110
pus of Blaise Pascal University in Clermont Ferrand. Since 2002, meteorological variables have
111
been measured at this site. More details concerning the COPDD observatory can be found on the
112
following website; http://wwwobs.univ-bpclermont.fr/SO/mesures/index.php.
113
3) CRA-P2OA
114
The P2OA observatory is the southern most site (Fig. 1a). It is located in the Midi-Pyren´ees
115
region and is composed of two sites from the Observatoire Midi Pyr´en´ees (OMP): the Atmo-
116
spheric Research Center (CRA) in Lannemezan (43.13◦N-0.369◦E at an altitude of 600 m), and
117
the “Pic du Midi” (43.13◦N-0.37◦E at an altitude of 2877 m). On this platform, various in situ
118
and remote sensing instruments continuously measure the atmospheric dynamics, surface energy
119
balance, radiation, chemistry, aerosols and atmospheric electricity. Here, we use only meteorolog-
120
ical observations at the CRA site which is a rural site located on a plateau in the foothills of the
121
Pyrenees. At the CRA site, standard meteorological observations have been collected since 1995.
122
More details on the P2OA observatory can be found on the website http://p2oa.aero.obs-mip.fr/.
123
Given the geographical position of the three observatories, various local processes, such as urban
124
heat islands, cloud cover, and mountain/plain breeze circulations, snow cover, and clouds have a
125
role to play concerning daily temperature anomalies.
126
b. Data used
127
In order to base our analysis on a common period with a uniform data format, data from the
128
Meteo-France standard weather station hosted by CRA-P2OA were used for this study. We em-
129
ployed hourly values concerning temperature and incoming shortwave radiation at 2 m, wind speed
130
and direction at 10 m and rainfall between 2003 and 2013. In the framework of the current study,
131
a similar quality control, homogenization and a combination of variables from various sources
132
as seen in the SIRTA-ReOBS were performed for the meteorological variables collected in the
133
C´ezeaux-COPDD and CRA-P2OA observatories.
134
To characterize the influence of large-scale circulations, we based our study mainly on daily
135
temperature anomalies at 2 m above the surface at the three observatories. To ensure that our
136
anomaly was not affected by seasonal variability in temperatures, it was defined by comparison
137
to the average of the current month. We defined the daily temperature anomalies (aT(j)) for the
138
day j by removing the 2003-2013 monthly mean temperature at each observatory. This can be
139
expressed through the following equation:
140
aT(j) =<T >j−<T >[m,2003−2013] (1)
where<T>jis the daily mean temperature for the day j, computed from the mean of hourly tem-
141
peratures. <T >[m,2003−2013]is the monthly mean temperature calculated over the entire period,
142
andmis the month represented numerically. For example,<T >[1,2003−2013]was the temperature
143
averaged over all the days in January across the period 2003-2013.
144
c. Large scale analysis diagnostics
145
1) WEATHER REGIMES
146
Weather regimes enable us to describe large-scale atmospheric circulations in a simple man-
147
ner. With this in mind, we used the classification of weather regimes used by Yiou and Nogaj
148
(2004) and based on the daily anomalies of sea level pressure (SLP) acquired from the NCEP (Na-
149
tional Center for Environmental Prediction) reanalyses (2.5◦×2.5◦) (Kalnay 1996). The weather
150
regimes are defined in the Euro-Atlantic region (80◦W-30◦E, 30-70◦N) (Fig. 1b, (the larger black
151
square one) and determined from the “K-means” algorithm, computed from the first 10 Empirical
152
Orthogonal Functions (EOFs) of seasonal SLP anomalies (Cheng and Wallace 1993; Michelan-
153
geli et al. 1995) from 1948 to 2014. The classification used in this study therefore depends on the
154
season. Figure 2 illustrates the four weather regimes defined in winter and their occurrence during
155
the 1948-2014 period. We note in this Figure, the positive (reg. 3) and negative (reg. 4) phases of
156
the North Atlantic Oscillation (respectively NAO+ and NAO−), a “Scandinavian blocking” (reg.
157
2), and the “Atlantic Ridge” (reg. 1). Weather regimes appear with a similar frequency with a 27%
158
occurrence for NAO+and “Scandinavian blocking”. During the transitional seasons of spring and
159
autumn, a classification into weather regimes is not always appropriate due to seasonal shifts (Vrac
160
et al. 2013). Vrac et al. (2013) found that spring frequently corresponds to an early summer or a
161
longer winter, and that autumn is related to a longer summer or earlier winter, making a definition
162
of a regime during these two seasons difficult. Here, we do not consider this classification for
163
transitional seasons. It is also necessary to consider the stability of the regimes during the winter
164
and summer, as they are sometimes not well defined, and only transitory.
165
In order to eliminate the days with ambiguous classification in winter and summer, we use a
166
criterion based on the Euclidean distance and the spatial correlation from the nearest weather
167
regime deduced by the K-means method. We filter the classification by eliminating the days for
168
which the Euclidean distance from the nearest weather regime is larger than 10 hPa and with a
169
spatial correlation with the nearest weather regime lower than 0.15. We eliminated 5.2% (52 days)
170
and 10,8% (109 days) of the total days in winter and summer respectively.
171
Here, we are interested in the influence of the large-scale atmospheric regimes on the variabil-
172
ity of daily temperature anomalies (equation 1) at the three observatories. Figure 3 shows the
173
box plot of daily temperature anomalies in winter and summer for each site and for each weather
174
regime during the 2003-2013 period. This Figure indicates that in winter, NAO+yields relatively
175
milder temperatures at all sites, while NAO− and blocking are characterized by relatively colder
176
temperatures at all sites. During Atlantic Ridge conditions, the occurrence of either warmer or
177
colder temperatures than those on average is relatively similar, except at SIRTA, where the winter
178
is mostly mild when this regime prevails. It is however, important to note that specific anomalies,
179
warm or cold, can occur whatever the weather regime at SIRTA, whilst very cold winter days are
180
unlikely to occur at C´ezeaux-COPDD or CRA-P2OA when NAO+ or Atlantic Ridge conditions
181
exist. Extreme temperature anomalies are more frequent at SIRTA, and variability is usually en-
182
hanced, except during NAO+. In summer, the weather regimes have almost the same effect at all
183
sites, even if the variability at C´ezeaux-COPDD is greater than at the other two sites, and extremes
184
are enhanced. At C´ezeaux-COPDD, the Atlantic Ridge and NAO+ have positive daily anomalies
185
on average in winter (0.6◦C and 0.2◦C respectively) and summer (0.6◦C and 1.9◦C respectively),
186
indicating mild and warm temperatures respectively during these two seasons. These results are
187
consistent with those of Yiou et al. (2007) in the fall/winter of 2006/2007. From these results, we
188
conclude that the weather regimes derived from the SLP data do not explain the daily temperature
189
anomalies at the three observatories in winter and summer.
190
2) LARGE-SCALE FLOW ANALOGUES
191
The slight difference in the anomaly of mean temperatures among the weather regimes in sum-
192
mer, the large variability in the daily temperature anomalies and the fact that the weather regimes
193
are not easily defined in spring and autumn, motivated us to augment the regime approach with
194
the flow-analogue method.
195
The method of atmospheric flow analogues was first introduced by Lorenz (1969). Since then, it
196
has found many applications, including weather prediction (Van den Dool 2007). Yiou et al. (2007)
197
used this approach to infer the connection between surface climate variables and atmospheric
198
circulation. In this study, we use the flow-analogue method developed by Yiou et al. (2007) and
199
used by Chiriaco et al. (2014) and Cattiaux et al. (2010) to study climate variability across Europe.
200
For each day during the eleven year period (2003-2013), we looked for days within the same
201
time series which had similar large-scale atmospheric conditions. For this, we considered field
202
anomalies of geopotential height at 500 hPa from the ERA-Interim (ERAI) reanalyses (0.75◦ ×
203
0.75◦) of European Center for Medium-Range Weather Forecasts (ECMWF) (Dee et al. 2011), a
204
typical diagnostic tool for large-scale circulations. Analogue days were found by minimizing a
205
Euclidean distance and maximizing a Spearmann correlation. More details on the flow analogues
206
method can be found in Yiou et al. (2007).
207
By using flow analogues to quantify the relative influence of the large, local, and mesoscale
208
processes on surface temperature anomalies, we considered two nested domains. The first domain
209
covers the Euro-Atlantic region (80◦W-30◦E, 30◦N-70◦N) (Fig. 1b, black square). This domain
210
is also the one used by Cattiaux et al. (2010); Vautard and Yiou (2009); Yiou et al. (2007) and
211
Chiriaco et al. (2014) to establish the link between extreme events (cold waves, heat waves and
212
drought) and large-scale conditions over Europe. The second domain covers the area 21◦W-30◦E,
213
30-60◦N (Fig. 1b, white square). Compared to the larger domain, this smaller domain (mesoscale)
214
weighs the influence of the Mediterranean sea on synoptic circulations more heavily than the
215
Atlantic Ocean.
216
For each day in our studied period and for each domain considered, we kept a maximum of ten
217
analogues, which satisfied the following two criteria: (i) the Spearman spatial correlation had to be
218
greater or equal to 0.6, ensuring the quality of the similarity, and (ii) they should not be closer than
219
6 days from the current day, in order to ensure that the analogues were independent of the target
220
day (assuming a decorrelation time of 3 days before and 5 days after the target day). These criteria
221
eliminated 6,4 % (around 256 days) of the days from the large domain and 3 % (around 119 days)
222
from the small domain. The scores are higher in winter, spring and autumn than in summer for
223
both domains. We found 134 and 54 unselected days respectively for the large and small domain
224
in Summer.
225
d. Analysis protocol
226
To quantify the contribution of local processes and large scale circulations at each site, we com-
227
pared the observed temperature anomalies to the temperature anomalies observed during the ana-
228
logue days. Figure 4 illustrates this approach for the year 2007 with analogues of circulation
229
computed over the small domain. It shows that, for all observatories, the analogues reproduced
230
the observed temperature anomalies quite well, but there was also a great variability between ana-
231
logues. For certain days, the analogues could not capture the amplitude of the observed anomalies,
232
as can be seen in the example from 17 to 20 January 2007 on all sites, February 2007 at SIRTA
233
and C´ezeaux-COPDD, at the end of August 2007 at CRA-P2OA, and at the end of April 2007 at
234
SIRTA. A smaller standard deviation of the ten anomalies of analogous days combined with an av-
235
erage closer to the temperature anomaly of the day in question means that the large scale explains
236
the anomaly. In our study, we investigated whether this departure from the observed series relative
237
to the envelope defined by the atmospheric conditions on analogue days can be explained by local
238
processes. Weather regimes were used to describe and better understand the large-scale influence
239
(see indications of the regimes in Fig. 4).
240
3. Analysis of large-scale conditions versus local processes
241
In order to estimate the influence of the Mediterranean Sea relative to the Atlantic Ocean at the
242
three sites, we first evaluated the ability of the analogues to represent the observed series using
243
the two different domains described above. Afterward, the difference between the observed series
244
and the temperature anomalies of the analogues was quantified by the definition and the use of
245
an anomaly index. Finally, we focused on specific periods during which the difference was larger
246
than 1.5 ◦C, tried to identify the relevant processes, and discussed the relative contribution of
247
large-scale and local processes.
248
a. Sensitivity to the Mediterranean Sea
249
Figure 5 presents the correlation between observed anomalies and those deduced from flow
250
analogues in the large and small domains (Fig. 1b) for each site and for each season. All observed
251
daily temperature anomalies for each season are correlated with those of their 10 analogue days.
252
Thus, for each season of each year from 2003 to 2013, we have one correlation coefficient. This
253
Figure points out larger correlation coefficients in the small domain than in the large domain. This
254
is obvious for the two southern observatories whatever the season, whereas higher correlation
255
coefficients across the small domain are observed only in summer and spring for SIRTA. This
256
shows that SIRTA is more influenced by large-scale air masses coming from the Atlantic than by
257
mesoscale processes induced by orography and the presence of the Mediterranean Sea, which can
258
strongly influence the weather across southern France (e.g Ducrocq et al. (2008)).
259
We found a large spatio-temporal variability in the correlation coefficients. CRA-P2OA indi-
260
cated, on average, the lowest correlation coefficients for the two domains (0.52 and 0.35 respec-
261
tively for the small and large domain) compared to the other two sites (for the small domain,
262
C´ezeaux-COPDD and SIRTA had respectively 0.55 and 0.57 and for the large domain, 0.38 and
263
0.46 respectively). This difference can be due to the fact that CRA-P2OA is located in proximity
264
to the Pyrenees, where local processes linked with topography exist: for example, local convec-
265
tion or plain-mountain breeze circulations are more frequent in summer. The two cases of very
266
low correlation with the large domain (at the bottom-left of each subplot in Fig. 5 with green and
267
black colors) were observed in autumn 2011 at each site, during the winter of 2006 at SIRTA, and
268
during the winter of 2008 at CRA-P2OA and C´ezeaux-COPDD. The autumn of 2011 was excep-
269
tionally warm. It was indeed the second warmest autumn during the period 1948-2011, after 2006,
270
according to Cattiaux and Yiou (2012). Cattiaux and Yiou (2012) found that the flow analogues
271
underestimated the amplitude of the seasonal temperature anomaly in Europe during this specific
272
season. This suggests that global warming plays an important role by increasing the concentration
273
of greenhouse gases: the advected air mass is warmer, but it can also enhance local feedbacks.
274
In the following section, we evaluate the flow analogues approach by considering only the
275
smaller domain, in order to quantify the influence of local processes on the climate variability
276
at the three sites.
277
b. Large-scale influence
278
We have attempted to better quantify the relative contribution of large scale versus local pro-
279
cesses on the amplitude of temperature anomalies. Since the average signal of the analogues have
280
inherently lower magnitude fluctuations, we introduce Im, a new normalized index to facilitate
281
comparison of the observations and analogue series.
282
This indexIm, defined for each monthm, represents the monthly average anomaly<aT(j)>m
283
relative to the standard deviation of the 2003-2013 daily anomalies for the given monthm. We
284
computeImwith:
285
Im= <aT(j)>m q
<(aT(j)−<aT(j)>[m,2003−2013])2>m
, (2)
where<aT(j)>[m,2003−2013]is the average anomaly of temperature of the current monthmduring
286
the period 2003-2013. In other words, the red line in Fig. 4 is averaged monthly and divided by the
287
standard deviation computed from the monthly anomalies for the period 2003-2013. Concerning
288
the analogue signal, the same definition of the index is applied using all observed anomalies in the
289
analogue days.
290
Figure 6 represents the time series of this index across the period 2003-2013. The flow analogues
291
reproduce the variability of surface temperature anomalies particularly well. The correlation coef-
292
ficients betweenImfor observations and analogues are 0.80 for SIRTA, 0.85 for C´ezeaux-COPDD
293
and 0.86 for CRA-P2OA. This means that the large scale actually plays a predominant role in
294
creating the temperature anomaly variability on monthly scales, which is not surprising.
295
In Fig. 6, one may note the spatio-temporal variability ofImat the three observatories. The years
296
2003 and 2011 were the warmest years at every site for the period 2003-2013, whereas the coldest
297
year at every site was that of 2010 with negativeImfor every month.
298
While the general trend is well captured by Im for analogue days, the magnitude of certain
299
events is not reproduced. For example, February 2007 was exceptionally warm with Im larger
300
than 1.7 at SIRTA and C´ezeaux-COPDD according to observations. This peak in temperature
301
is not reproduced by flow analogues with an index of around 0.5 (Fig. 6) when using the small
302
domain, and is even negative when using the large domain (not shown). The large anomaly is not
303
observed at CRA-P2OA. The spatial variability of temperature anomalies during this winter and
304
the difference between observed anomalies and analogues allow us to hypothesize that specific
305
synoptic-scale features leading to local anomalies that are not resolved by the analogue approach
306
alone and that local processes may have played a specific role at each site during this period.
307
c. Analysis of specific events during winter 2007
308
We focused on the winter of 2006/2007 in order to further investigate the contribution of large
309
and local-scale processes on the spatio-temporal variability of daily temperature anomalies at the
310
observatories. Note that winter 2007 appears to be the warmest of our study period: it was the sec-
311
ond warmest winter in France since 1959 according to climatology established by Meteo-France
312
(http://www.meteofrance.fr/).
313
Figure 7 shows the time series of daily temperature anomalies for the winter of 2006/2007
314
from observations and analogues. It focuses in on the period from January to February 2007 in
315
Fig. 4. During this period, two regimes, “NAO+” and ”Atlantic ridge” are persistent. “NAO+” is
316
associated with a southwesterly flow over Northern Europe (Michelangeli et al. 1995). We showed
317
in section 2c that the two regimes, “NAO+” and “Atlantic Ridge” are usually the warmest in winter
318
at all three sites. These results are consistent with those of Yiou et al. (2007) for the exceptionally
319
warm 2006/2007 fall/winter. The regime NAO−appears between 22 and 26 January, with negative
320
anomalies at all sites. Snow was observed at SIRTA on 23 January, and from 23 to 25 January at
321
CRA-P2OA and C´ezeaux-COPDD.
322
We focused on specific warm events during the winter of 2007 to investigate the role of lo-
323
cal processes on the spatio-temporal variability of daily temperature anomalies. We will further
324
analyze two periods /dates: the period 17 to 19 January and a single day: 16 February 2007.
325
1) 17-19 JANUARY 2007CASE
326
The period from 17 to 19 January encompasses the warmest anomalies of the month of January
327
2007 at all sites (Fig. 7) with spatial variability in the amplitude: the southern site (CRA-P2OA)
328
shows the lowest daily temperature anomalies compared to the other two sites (warmest anomaly
329
of 9.4, 10 and 7.2 ◦C at SIRTA, C´ezeaux-COPDD and CRA-P2OA respectively, on 18 January
330
2007). The observed positive temperature anomalies are higher than those of analogue days for
331
the whole 17-19 January period at SIRTA and C´ezeaux-COPDD and only at CRA-P2OA for the 18
332
January. Despite the spatial variability in the temperature anomalies, 18 January 2007 indicates an
333
anomaly on a large scale and one can wonder why the anomaly’s amplitude of such a large-scale
334
event is not reproduced by any of the analogue days.
335
To answer this question, the large-scale meteorological situation of analogue days is verified
336
using satellites and ERAI reanalyses and local effects are analyzed based on the meteorological
337
history of the surface measurements and radiosoundings. The meteorological history provides a
338
view of the atmospheric conditions of previous days on a local scale. We consider, therefore,
339
the diurnal cycles on 18 January 2007 and on the two previous days (16 and 17 January 2007)
340
to point out the effect of the ‘‘local meteorological history’’ at each site. Similarly, the ‘‘local
341
meteorological history’’ of the five best analogue days on 18 January 2007 is presented. For
342
example, if 20 December 2011 is one analogue day for 18 January 2007, the time series from 18
343
to 20 December 2011 are displayed.
344
The large-scale circulation is the NAO+ regime on 17 and 18 January, and ‘‘Atlantic ridge’’
345
on 19 January (Fig. 7). Figure 8 shows the wind speed and direction at 600 hPa from the ERAI
346
reanalyses. It indicates an increasing westerly wind over France during the 16-18 January period.
347
The five most accurate analogues are generally similar with an increasing wind speed from 17
348
to 18 January and show similar wind directions. On 16-17 January, only one analogue indicates
349
similar wind speed and direction. However, wind speed varies from one analogue to another.
350
The reflectance in the visible channel at 0.6µm of the MeteoSat Second Generation (MSG) at
351
1300 UTC on 18 January 2007 is shown in Fig. 9a. Significant cloud cover over the North Atlantic
352
and Europe was observed with a window of clear sky over the Mediterranean basin, the South of
353
Spain and the Pyrenees region. Similar cloud cover was also observed on 17 and 19 January
354
2007 (not shown). The method of Wang and Rossow (1995) was applied to the vertical profile of
355
relative humidity from the radiosoundings at Trappes on 18 January (Fig. 9b) to define the cloud
356
base height. Wang and Rossow (1995) used among other criteria, 87% and 84% as maximum and
357
minimum relative humidity thresholds respectively and relative humidity jumps exceeding 3% at
358
cloud-layer top and base to characterize a cloud-layer. With this method, we find in Fig. 9b that
359
on the 18 January, cloud cover was dominated by low-level clouds with a base not exceeding 700
360
m in height at 1100 UTC. At 2300 UTC, cloud cover descended and thickened. Based on the
361
Meteo-France weather service station, drizzle was observed that night, with 0.8 mm falling at this
362
site. Combining the satellite image and vertical profiles of relative humidity, we find that these low
363
clouds were stratocumulus clouds associated with the stable atmospheric conditions in Southern
364
Europe linked to the NAO+ regime. Indeed, the stratocumulus clouds occur widely over Europe
365
in January, according to Hahn and Warren (2007). A similar analysis of the vertical profiles of
366
relative humidity for the five most accurate analogue days (Fig. 9b) shows that 18 January 2007
367
was the cloudiest day: either there was no cloud cover (analogue day number 3), or there was cloud
368
cover which disappeared between 1100 and 2300 UTC (analogue day number 1), or much thinner
369
cloud cover (analogue days number 4 and 5). We can expect an effect due to this cloud layer on
370
18 January since it impacts the radiative budget at the surface at SIRTA and C´ezeaux-COPDD.
371
The meteorological history of 18 January and its five most precise analogue days was analyzed
372
with surface measurements. The large scale cloud cover, discussed previously (Fig. 9), impacts
373
incoming solar radiation (ISR) (Fig. 10). The ISR measured at the surface increases from north
374
to south, with very cloudy conditions at SIRTA for every day and almost no reduction of ISR
375
at CRA-P2OA. The integration of ISR across the three days defining the meteorological history
376
period (not shown) demonstrates low levels of ISR for the observed days compared to the analogue
377
days at SIRTA and C´ezeaux-COPDD, contrary to CRA-P2OA.
378
Figure 10 presents the time series of temperature, incoming shortwave radiation at 2 m, wind
379
speed and direction at 10 m above the ground. Cloud cover also clearly impacts the diurnal cycle
380
of 2m-temperature (Fig. 10); Low-level cloud cover at SIRTA reduces the cooling of the earth and
381
damps the diurnal temperature cycle. This is also the case at C´ezeaux-COPDD, on 18 January.
382
On the contrary, a large diurnal temperature cycle can be observed at CRA-P2OA on most of the
383
days, especially during the period 16-18 January.
384
At SIRTA, the westerly wind direction at the surface is consistent with the synoptic wind (Fig. 8).
385
The wind direction at C´ezeaux-COPDD is quite variable but maintains a westerly direction on av-
386
erage, whereas a clear effect of the mountain range can be observed on 16, 17 and some of 18
387
January at CRA-P2OA, with some north-easterly slope winds during the day and southerly at
388
night, a sign of the plain-mountain diurnal circulation. Figure 10 shows the diversity of the con-
389
ditions observed during the analogue days at CRA-P2OA, which makes the comparison difficult.
390
Among the five most accurate analogue days, only three are cloud-free. All of them indicate a
391
reversal of the wind direction twice a day. This is characteristic of the slope wind, which seems
392
to play an important role and blurs the comparison of the diurnal cycle. During winter, a lack of
393
cloud cover may allow weak convection over mountains, and certainly greater radiative cooling
394
at night. This southerly mountain breeze during the night advects cool air from the mountains
395
and is associated with low temperatures at night. The mountain breeze which occurs during the
396
NAO+ regime could then reduce the positive temperature anomaly tendency associated with this
397
regime. The meteorological history of 18 January shows a slope wind regime until noon, when a
398
clear westerly wind settles at the surface. From that moment, the temperature clearly increases,
399
and remains high during the night, with no mountain breezes, between 18 and 19 January. The
400
daily mean temperature then leads to a larger positive temperature anomaly compared with the
401
analogue days with slope winds lasting all day.
402
From these large and local-scale analyses of the observed days and their analogue days, we
403
can ascribe this positive temperature anomaly to a large-scale event observed at the three sites.
404
The NAO+ regime, which advects mild temperatures from the Atlantic ocean, is characterized
405
by the warmest temperature anomaly in winter (Fig. 3). The flow analogue method shows some
406
limitations, however, in representing this event. The cloud layer is particularly low and deep,
407
and lasts for three days over the northern part of France, whereas nothing in the meteorological
408
history of the analogue days indicates such conditions. This cloud cover could imply a warming
409
radiative effect over SIRTA and C´ezeaux-COPDD during the 17-19 January period, which would
410
amplify the positive anomaly due to what is already mild air advection. While the low cloud
411
cover observed during this event is not a local effect, its radiative interaction with the surface is
412
dependent on surface temperature and can be considered a local effect.
413
In conclusion, it seems that this abnormal warm event stands out from the analogue days, for
414
various reasons at SIRTA and C´ezeaux-COPDD in the first instance and then at CRA-P2OA.
415
The large-scale positive anomaly associated with the NAO+ regime is amplified at SIRTA and
416
C´ezeaux-COPDD by the warming radiative effect of an unusually low cloud cover occurring over
417
the two sites during the 11 year period. This event, lasting for three days, lead to warmer anomalies
418
than on analogue days. Meanwhile, in CRA-P2OA, the absence of clouds lead to a down-valley
419
wind regime which tends to cool the air at night and to reduce the NAO+ regime warm anomaly.
420
The down-valley wind was observed on 18 January until midday and did not occur the following
421
night. This led to higher nocturnal temperatures and warmer daily temperature anomalies than on
422
analogue days the following night. These results show that radiation and cloud cover are important
423
predictors of daily temperature anomalies in winter at this observatory.
424
2) 16 FEBRUARY2007CASE
425
16 February 2007 is an example of a case where the temperature anomaly largely exceeded the
426
range of the flow analogues at a single site. A strong and warm anomaly of 12.3◦C was observed
427
at CRA-P2OA on that day, while all analogues showed an anomaly below 8◦C (Fig. 7). At the two
428
other observatories, the temperature anomaly on this day was within the envelope of the analogues.
429
The synoptic atmospheric conditions on 16 February were forced by the presence of very low
430
pressure centered over Iceland, and its associated thalweg extending from the island towards the
431
south, near Spain and Morocco. This situation, which often announces the arrival of a front,
432
generated a south-southwesterly wind regime in altitude, bringing dry and warm air from the
433
south. The wind at 600 hPa across the three sites and deduced from the reanalyses of the European
434
Center is shown in Fig. 11 for 16 February and for its analogous days. The analogues have the
435
same types of southwesterly wind regime across the three sites. This situation generally leads to a
436
positive temperature anomaly due to the southern origin of the air mass in many such cases. For
437
this reason, on average, the envelope of the analogues shows a positive temperature anomaly at all
438
sites (Fig. 7).
439
Southerly winds over the ridge of the Pyrenees correspond to the typical situation of the so-
440
called foehn phenomenon: the east-west orientated mountain ridge is an obstacle for the flow,
441
which can be partially blocked in the lower layers and which can bypass the ridge, with the flow
442
splitting at its sides, or/and passing over and through it across the mountain passes (Scorer 1949,
443
1953, 1955; Scorer and Klieforth 1959; Seibert 1990; ´Olafsson and Bougeault 1997; Jiang et al.
444
2005). The adiabatic descent of air in the lee, usually occurring together with the flow over the
445
mountain, is associated with a typical drying and warming in the lower lee air layers on the French
446
side (‘foehn effect’).
447
One of the most important governing variables for this phenomenon is the upwind wind profile,
448
and particularly the upwind component, which is perpendicular to the ridge: the larger this com-
449
ponent, the easier it is for the flow to go over the mountain and generate the foehn effect (Seibert
450
1990). For the Pyrenees in the vicinity of the CRA-P2OA site, we evaluate the cross-component
451
at 210◦ azimuth (±10◦): that is, a wind with this direction (which is very similar to a southerly
452
wind) travels exactly transversely to the ridge, on a 150 km horizontal scale. This direction is
453
also aligned with the main Aure Valley, which is situated south of the CRA-P2OA observatory
454
and North-South orientated. Figure 12a shows the upwind profiles of the cross-ridge component
455
(projection of the wind on the 210◦ axis), for 16 February and for all analogues, at 0000 UTC.
456
These are deduced from the radiosoundings launched daily from Zaragosa in Spain. Zaragosa is
457
located about 150 km south of the ridge of the Pyrenees, and 200 km from the CRA-P2OA site.
458
These profiles confirm the potential to generate the foehn effect at CRA-P2OA for most of the days
459
shown, as this component is positive for most cases above 1000 m. It also reveals that 16 February
460
is the case with the strongest 210◦ upwind component between 1000 m and 6000 m, especially
461
below 3500 m, making it the most favorable case for a strong foehn event (the highest peak in the
462
Pyrenees is at 3400 m). Figure 11 also shows a significant increased in wind speed upwind of the
463
ridge during the day.
464
Figure 12b is a the visible image of MSG at 1500 UTC on 16 February 2007. This day was
465
marked by large cloud cover over the western Atlantic and northern Europe, and a clear sky above
466
the Mediterranean basin and eastern Europe. Cloud cover over the Pyrenees shows that the sky
467
was clear in Spain and in the lee of the mountain (where CRA-P2OA is situated). Farther to the
468
north, a cloud with a well-defined southern border, typical of the upward branch of a mountain
469
wave, can be observed, and is usually associated with foehn and southerly overpassing flows. The
470
clear sky in Spain reveals a ‘‘dry foehn’’ as opposed to some cases, where clouds are blocked on
471
the Spanish side, with some rain that can contribute to the drying and warming effect of the air in
472
the lee on the French side. This means that on 16 February, air mass was generally very dry at the
473
large scale, a fact which is confirmed by the radiosoundings taken at Zaragosa, Bordeaux (Atlantic
474
French coast), Trappes (close to Paris), and the synoptic situation discussed before. The Trappes
475
soundings at 1100 and 2300 UTC show very dry and warm air between 800 m and 4000 m. Above
476
this altitude, fine medium clouds (of about 500 m) are observed (not shown).
477
We can now consider the observations at the surface of the different observatories. Figure 13
478
presents the evolution of the meteorological variables observed close to the surface on 16 February
479
2007, and its analogues. The most striking feature is found in the surface wind at CRA-P2OA: for
480
all the analogues, the wind at the surface is southerly during the night and northerly during the day.
481
This, along with the low associated wind speed (below 6 m s−1), is indicative of the mountain-plain
482
diurnal circulation. That is to say, although the upwind flow is from the south, and sometimes has
483
a strong wind speed (Fig. 12a), this does not prevent the plain-mountain circulation from setting
484
up during those analogue days. It is actually quite classic, with the southerly wind kept at a higher
485
level. Note that this does not prevent the foehn effect (warming and drying in the lee), or the
486
warmer local temperature that can be found at this site relative to the other sites. On 16 February,
487
however, the wind at the surface of CRA-P2OA remained southerly all day, with the wind speed
488
increasing during the day, by up to 10 m s−1 at times. This means that for this specific day, the
489
upwind flow was strong enough to be able to create a downslope wind throughout the entire day in
490
which case, the warming and drying effect in the lee is still larger. This is consistent with Fig. 12a,
491
which shows the characteristics of this day in terms of upwind conditions. It probably explains
492
most of the temperature anomalies found at CRA-P2OA, which exceed the usual anomalies found
493
in analogous synoptic situations (Fig. 7).
494
At C´ezeaux-COPDD, this synoptic situation does not lead to a marked anomaly, but the general
495
dry and warm air leads to a large diurnal increase. The night of 16 to 17 February may have been
496
influenced by a small foehn effect, in the presence of westerly winds (typically occurring in the
497
‘‘Chaine des puys’’ mountains to the west of the site). The air temperature does not decrease
498
much, and the wind continues to arrive from the west. At SIRTA, the wind at the surface is
499
easterly, surprisingly, while the sounding at Trappes shows a strong southerly flow down to the
500
lowest levels of the atmosphere. This weak easterly wind at SIRTA seems unconnected to the
501
warm, dry southerly air above, and could explain the relative lower temperature found on 16
502
February (relative to its analogues).
503
This event shows how meso-β scale processes linked with orography can amplify a temperature
504
anomaly which is primarily forced at the synoptic scale. This specific type of amplification has
505
been previously observed by Takane and Kusaka (2011) in Japan in the summer.
506
4. Summary and Conclusions
507
This study aimed to evaluate the relative contribution of large-scale atmospheric circulation and
508
more local processes to daily temperature anomalies over a north-south transect of France. The
509
study was based on the observations of meteorological variables at three observatories and on
510
NCEP and ECMWF reanalyses. The flow analogues method was used in particular to diagnose
511
the fingerprint of the large-scale synoptic circulations concerning the temperature anomaly, and to
512
highlight the potential role of local processes in inhibiting or amplifying the anomaly.
513
The analysis of weather regimes over the Euro-Atlantic region shows that the large-scale atmo-
514
spheric circulations have an important influence on the daily temperature anomalies at the three
515
observatories in winter. The “NAO+” and “Atlantic Ridge” appear to be the warmest regimes in
516
this season. While the influence of the four weather regimes on daily temperature anomalies does
517
not statistically differ at the three observatories in the summer due to strong variability within each
518
regime, extreme anomalies are associated with one or two regimes at all observatories except for
519
that of SIRTA.
520
The flow analogue approach applied over two different domains shows that SIRTA is less af-
521
fected by the mesoscale processes formed around the Mediterranean Sea than the other two obser-
522
vatories, which is not surprising considering its northern location.
523
The atmospheric circulation analogue method demonstrates the large correlation between a
524
monthly temperature anomaly index calculated from the observed series and that which is pro-
525
vided by the representation of the analogues. This highlights the predominant role played by the
526
large-scale situation in the temperature anomalies. Sometimes, however, the amplitude of the
527
monthly temperature index is not captured by the flow analogues and shows a large spatial vari-
528
ability between the three observatories. It is suggested that these discrepancies are related to local
529
processes. Two specific events revealed in the warmest winter in the period 2003-2013 are further
530
analyzed to test this hypothesis: 1) the 17-19 January 2007 event which had the strongest posi-
531
tive temperature anomaly at the two northern observatories (SIRTA and C´ezeaux-COPDD) and,
532
2) the 16 February 2007, for which only CRA-P2OA indicated a very large positive temperature
533
anomaly, found beyond the signal of the set of analogues.
534
From the analysis of these two events, the impact of several local processes have been identified:
535
1) the local impact of non-local cloud cover during westerly wind conditions in winter: low-level
536
clouds have been shown to increase the positive temperature anomaly at SIRTA and C´ezeaux-
537
COPDD in these conditions, partly due to the positive radiative green-house effect of the clouds.
538
2) the orographic impact: CRA-P2OA and C´ezeaux-COPDD are both in proximity to mountains
539
and are frequently impacted by either foehns or slope wind effects. In a weak large-scale situation,
540
the slope breeze easily settles at CRA-P2OA, and can transport cool air from the mountains during
541
the night in winter. Foehn events observed at both the CRA-P2OA and C´ezeaux-COPDD sites with
542
southerly and westerly wind conditions respectively, can amplify positive temperature anomalies,
543
originally forced by large scales.
544
The analysis of two specific events reveals that some local processes are able to modulate the
545
trend of the daily temperature anomaly driven by the large-scale atmospheric circulation. How-
546
ever, such a phenomenological approach remains difficult, since the understanding of one event
547
necessitates the analysis of the meteorological history of not only the event itself, but also of its
548
analogue days. To investigate the impact of local processes, a systematic study of all cases in
549
which observations differ from analogue days would be necessary.
550
Departures between observed local anomalies and analogues might not only be due to local
551
processes but also to differences between the observed event and its analogues on the synoptic
552
scale, which would not be adequately resolved by the classical analogue approach employed. A
553
possibility for the investigation of this is the combining of different variables in the analogues
554
method (vorticity, water vapor, temperature, wind). Even if this would require much longer series
555
in order to ensure a large enough number of analogues for each day.
556
Acknowledgments. This work was carried out in the context of ROSEA, and funded by AL-
557
LENVI. The ROSEA program now belongs to a larger program and national network called AT-
558
MOS (Atmospheric Short-Lived Climate Forcers Observing System). The administrative and
559
technical supervision of the observatories have been acknowledged. Part of the data used here
560
were collected at the Pyrenean Platform for Observation of the Atmosphere P2OA, Observa-
561
toire de Physique du Globe de Clermont Ferrand OPGC and Site Instrumental de Recherche par
562
T´el´ed´etection Atmosph´erique SIRTA. P2OA facilities and staff were funded and supported by the
563
Observatoire Midi-Pyr´en´ees (University of Toulouse, France) and the CNRS (Centre National de
564
la Recherche Scientifique) INSU (Institut National des Sciences de l’Univers). OPGC facilities
565
and staff were funded and supported by the Blaise Pascal University of Clermont Ferrand and
566
CNRS (Centre National de la Recherche Scientifique) INSU. We are grateful to NCEP, ECMWF,
567
and Meteo-France for providing the observation data and global model reanalyses used in this
568
study. We acknowledge the CNES for partially funding M. Chiriaco’s research. P. Yiou was sup-
569
ported by an ERC advanced grant (No. 338965 - A2C2 ). The authors would like to thank the three
570
anonymous reviewers for their fruitful comments, which helped us to improve the manuscript. Fi-
571
nally, we thank Naomi RIVIERE and Eric PARDYJAK for carefully proof reading our manuscript.
572
References
573
Bastin, S., M. Chiriaco, and P. Drobinski, 2016: Control of radiation and evaporation on tem-
574
perature variability in a wrf regional climate simulation: comparison with colocated long term
575
ground based observations near paris.Climate Dyn., doi:10.1007/s00382-016-2974-1.
576
Bo´e, J., and L. J. Terray, 2014: Land-sea contrast, soil-atmosphere interactions and cloud-
577
temperature interactions: interplays and roles in future summer european climate change.Cli-
578
mate Dyn.,42(3-4), 683–699.
579